1

1-x xInfrared characterization of a-Si:H/a-Si C :H interfaces

J. Bertomeu, J. Puigdollers, J.M. Asensi and J. Andreu.

Laboratori de Física de Capes fines (LCFC), Departament de Física Aplicada i Electrònica,

Universitat de Barcelona. Av. Diagonal, 647. 08028-Barcelona (Catalonia, Spain).

E-mail: jberto@electra.fae.ub.es

Fax: 34-3-402 11 38

Phone: 34-3-402 11 44

Abstract

1-x xInfrared spectroscopy was used to characterize three series of a-Si:H/a-Si C :H

multilayers in which their geometrical parameters were varied. The infrared active vibrational

groups in their spectra and the interference fringes in their absorption-free zone were studied to

analyze the interfaces and the changes that are produced in very thin layers. Our results show that

hydrogen is bonded to silicon only in monohydride groups. No additional hydrogen could be

1-x xdetected at these interfaces. The deposition of very thin a-Si C :H layers seems to affect their

porous structure, making them denser.

2

1. Introduction

Amorphous silicon-based multilayers have been extensively studied during the last

decade. In addition to their intrinsic interest because of their properties and their claimed and

discussed quantum effects and their possible application to devices, these materials provide a

powerful tool to study the interfaces between amorphous silicon alloys.

One of the more interesting types of interface is that between hydrogenated amorphous

1-x xsilicon (a-Si:H) and hydrogenated amorphous silicon-carbon alloys (a-Si C :H). This interface

1-x xis very common in a-Si:H solar cells where a-Si C :H is used as a window layer because of its

wide bandgap, which enhances the spectral response at lower wavelengths [1].

1-x xThe H in a-Si:H/a-Si C :H multilayers has not been extensively studied and, except for

a paper from Yoshimoto and Matsunami [2], only casual references to the presence of additional

H at the interfaces [3,4] are reported. As far as H bonding is concerned, the identification of the

H bonding state is not straightforward due to the overlap of several contributions in the Si-H

nstretching band. In particular, the contribution around 2070-2090 cm can be attributed to Si-H-1

(n>1) groups [5], to Si-H groups on inner surfaces [6] or to Si-H bonds in the proximity of C

atoms [7].

In addition to the H incorporation scheme in these structures, another important issue is

whether the thicknesses of the layers at the nanometer scale could modify the structure of the

same material in homogeneous and thicker films (in the micron scale). This is particularly

1-x xinteresting in the case of a-Si C :H films which have been reported to present a porous structure

when their carbon content is high (i.e. when their bandgap is wider and, consequently, more

interesting optical properties are presented) [8,9].

In this paper we present an exhaustive study on these interfaces by using infrared

3

1-x xspectroscopy (IRS). Three series of a-Si:H/a-Si C :H multilayers with the same base materials

were used to study the H bonded in these interfaces.

2. Experimental procedures

2.1. Sample preparation

1-x xThree series of a-Si:H/a-Si C :H multilayers were deposited in an RF capacitively

coupled reactor with an automated substrate holder, as described elsewhere [10]. All the samples

were prepared at a substrate temperature of 300/C and an RF (13.56 MHz) power of 5 W

(27 mW cm ). The gas pressure was maintained at 20 Pa, and pure silane (20 sccm) and a-2

mixture of silane (1 sccm) and methane (19 sccm) were respectively used in a-Si:H and

1-x xa-Si C :H deposition. The plasma was maintained during the gas switching process, but

substrates were removed from plasma influence until conditions were stable [10]. The carbon

1-x xcontent (x) in a-Si C :H alloys was found to be 0.36 by X-ray photoelectron spectroscopy

(XPS).

Two series with variable mean composition (labeled A and B) and a series with constant

mean composition (labeled C) were prepared. The number of elementary layers was varied, but

the thickness of the multilayers was always kept around 0.5 :m. In series A, the a-Si:H layer

Si 1-x x SiCthickness (d ) was varied and the a-Si C :H layer thickness (d ) was kept at 37 Å, and in series

SiC SiB d was changed by keeping d at 55 Å. In series C, both thicknesses were changed by the

same ratio, and total thickness of the multilayers was kept constant. In Table I, we present the

thicknesses and number of elementary pairs of layers for the samples studied. The structure of

the multilayers was studied in a previous paper [10], showing a very good periodicity, but non-

4

(1)

strictly abrupt interfaces with a thickness of around 10 Å. The samples were deposited onto

intrinsic crystalline silicon wafers.

2.2. IRS characterization and data analysis

These multilayers and a homogeneous film of each of the two materials used in

multilayers were characterized by Infrared Spectroscopy in a BOMEM DA3 spectrometer in the

range between 400 and 4000 cm . The spectra were acquired in the normal incidence geometry,-1

under vacuum, and with a 4 cm resolution.-1

In order to obtain the absorption spectra of the samples, we measured the transmittance

Sof the substrate (T (<)) with the residual atmosphere spectrum as reference, the transmittance of

F/Vthe film/substrate system with the residual atmosphere spectrum as reference (T (<)), and the

F/Stransmittance of the film/substrate system with the substrate spectrum as reference (T (<)). The

F/S Sspectrum T (<) was divided by T (<) to eliminate the absorption contribution due to the

substrate. The resulting spectrum was then normalized by multiplying it by the mean value of the

ST (<) in the zone where no absorption bands are present. The result of this treatment (T(<)) is

F/Vcompared to T (<) and should be nearly identical except for the zone where there are absorption

bands due to substrate. The absorption spectra were obtained by fitting the zones in T(<) that did

not present absorption bands with the theoretical expression corresponding to the transmittance

of a non-absorbing film with an index of refraction n (which is considered constant in this range),

a thickness d, onto a semi-infinite substrate with an index of refraction s:

5

(2)

(3)

where: A = 16 n2

B = (n + 1) (n + s )3 2

C = 2 (n - 1) (n - s )2 2 2

D = (n -1) (n - s )3 2

These fits allow us to obtain the values of the parameters n, d and s. From these values

the absorption spectra ("(<)) are calculated from the T(<) expression:

where x = e , and A, B, C and D are as previously calculated from the fits.-"d

The H content bonded to silicon and carbon atoms in our samples was calculated from

n nthe integration of the Si-H and CH stretching bands:

nwhere K was taken as 1.4 × 10 cm for the Si-H stretching band and 1.0 × 10 cm for the20 -2 21 -2

nCH stretching band [1]. Although these K values are arbitrary and no calibration was performed

to our films with other techniques they are orientative and useful for comparison purposes.

3. Results and discussion

The T(<) spectra corresponding to the base materials used in the multilayers are presented

in figure 1. The a-Si:H spectrum shows only two significant bands, the Si-H stretching one

(centered at 2000 cm ) and the rocking one (centered at 630 cm ). The absence of contribution-1 -1

naround 2090 cm in the stretching band suggests that no Si-H (n>1) bonds are present and that-1

1-x xthe material is dense, without microvoids. The a-Si C :H film presents other characteristic bands

6

3 3of this material: the CH stretching band (2890-2940 cm ), the Si-CH bending band (1250 cm ),-1 -1

3 3the CH bending band (1350-1400 cm ) and the rocking and wagging bands of the Si-CH group-1

n(780 cm ). In addition, the stretching Si-H band is located around 2090 cm , without any-1 -1

significant contribution at 2000 cm . The absence of significant bands in the range 860-890 cm-1 -1

2suggests that no Si-H groups are present. Thus the contribution at 2090 cm should be due to-1

Si-H bonds in microvoids or to Si-H bonds with C atoms in their vicinity. The higher absorption

1-x xof this band in a-Si C :H layer is due to the higher porosity of this material, as deduced from

1-x xpreviously published results on the thermal desorption spectroscopy of a-Si C :H films with a

similar C content [8].

The fit of the spectra in the absorption-free zone for these samples gives a refraction

1-x xindex of 2.99 for a-Si:H and 1.75 for a-Si C :H. From the "(<) spectra we have calculated the

nH content bonded to silicon through equation (3) applied to the Si-H stretching band, and 4.7

1-x x× 10 cm and 1.2 × 10 cm were respectively obtained for a-Si:H and a-Si C :H films. The21 -3 22 -3

H content bonded to carbon atoms in the alloy was estimated to be 7.4 × 10 cm by the same21 -3

3method applied to the CH band. This amount is around 40 % of the H present in the alloy, which

is nearly the same percentage as that found for carbon content. This suggests that H has no

significant preference for Si or C atoms.

The T(<) spectra corresponding to the series A and B, both with variable mean

composition, are presented in figures 2 and 3. The first feature that stands out in these spectra is

the shape of the absorption-free transmittance spectra. The samples which presented a lower

mean carbon content showed softer interference fringes due to the smaller difference between

the mean index of refraction of the multilayer and that of the silicon substrate. This mean index

of refraction can be calculated through the following expression [11]:

7

(4)

(5)

The absence of any noticeable absorption band in the range 860-890 cm (corresponding-1

2to Si-H bending band) suggests that no Si atoms are bonded to more than one H atom. This leads

to the conclusion that all the 2090 cm contribution in Si-H stretching band is due to H atoms-1

on the inner surfaces of microvoids or to Si-H bonds in the neighbourhood of C atoms (in the

1-x xa-Si C :H alloy or in the a-Si:H layers near the interfaces). From the results corresponding to

the reference materials, we can conclude that the transmittance spectra of these multilayers are

between those of the reference materials, and that the multilayer structure does not significantly

influence the spectra. The calculation of the content of H bonded to Si atoms in these multilayers

through equation (3) shows that this depends only on the mean composition of the multilayer.

The H content as a function of the mean C content is plotted in figure 4. This mean C content

was calculated from the nominal thicknesses of the elementary layers and assuming the same C

1-x xcontent in the a-Si C :H layers that in homogeneous films. The H contents of the multilayers

are distributed along the line that connects the contents of the reference materials. Similar

behavior was observed for the H bonded to C atoms, but the scatter of the data is much higher

due to the lower absorption of the CH stretching band and the difficulty in the accurate

determination of the "(<) spectra from the transmittance spectra.

In order to perform a more detailed analysis of the possible influence of the multilayer

structure on the absorption spectra, we have synthesized the expected absorption spectra from

the experimental spectra of the reference materials through the following equation [11]:

The comparison between the synthesized spectra and the experimental ones did not show

8

significant differences in the shape of the curves. Nevertheless, a slight increase in the integrated

area of the experimental spectra of all the samples was observed, but no correlation between this

increase and the number of interfaces in the multilayers was found. This suggests that the amount

of additional H atoms which could be present at the interfaces should be very small. In fact, some

authors have estimated this amount to be of the order of 10 cm [3]. This amount is, in our14 -2

opinion, too small to be detected in our multilayers. In the sample A6, which presents the highest

number of elementary pairs of layers, this amount would mean an additional density of H atoms

of around 4 × 10 cm , which is less than 5% of the total H density.20 -3

0The effective index of refraction of the multilayers obtained from the fit of T (<) can be

used to deduce the index of refraction of the reference materials, by fitting equation (4). In figure

5 the effective index for the multilayers of series A is presented as a function of the varying

Si SiCthickness. The data were fitted by using n and n as parameters, and the best fit is also plotted

Si SiC Siin this figure. The values of these parameters were 2.99 for n and 1.84 for n . The n value is

SiCwithin the experimental error equal to that obtained for a a-Si:H layer (2.99), whereas that of n

1-x xobtained from the fit significantly differs from the a-Si C :H layer used as reference (1.75).

1-x xThis difference could be due to the lower incorporation of C or H in the a-Si C :H layers when

1-x xthey are in a multilayer structure, or to a reduction of the microvoids in the a-Si C :H when the

thickness of the layers is very small (37 Å in this series). Both explanations would imply an

1-x xincrease in the density of a-Si C :H and, consequently, an increase in the refraction index.

However, the H content in the multilayers is nearly the same as that corresponding to a

homogeneous film with identical mean composition as deduced from figure 4, and no differences

in the C incorporation with respect to homogeneous films are expected due to the multilayer

1-x xdeposition process, where self bias voltage during the deposition of a-Si C :H layers was

identical to the one measured during deposition of homogeneous films [10]. In our opinion, these

9

1-x xfacts suggest that the change of the microstructure of the a-Si C :H when it is arranged in a very

thin layer is the most plaussible explanation for the increase of the density. This change could

1-x xbe due to the thinness of the a-Si C :H layers, which prevents the growth of large microvoids.

The growth of a-Si:H layers on top them partially fills these growing microvoids, giving non-

abrupt interfaces, but also a low density of dangling bonds [12] and a higher density of the

1-x xa-Si C :H layers as a result.

Clarification about the presence of additional H at the interfaces could be expected from

our experiments on series C, where the mean composition was kept constant, and only the

thicknesses and number of elementary pairs of layers were varied. The transmittance spectra of

the multilayers of series C are presented in figure 6. The first feature that should be noticed is that

the shape of the absorption-free zone in the spectra is identical in all samples, suggesting that the

effective refraction index does not change. This could be expected from equation 4. The H

bonded to C atoms does not change when the number of interfaces increases. The most important

difference among these spectra comes from the Si-H stretching band. The 2090 cm contribution-1

increases when the thicknesses of the elementary layers decrease, whereas the total amount of

H in this band does not change significantly. This fact is more evident in the absorption spectra

of these samples, which is presented in figure 7. There is no systematic increase in the H content

in the samples to suggest the presence of additional H at the interfaces. The increase in the 2090

cm contribution is due to the fact that the density of Si-H bonds in the vicinity of C atoms-1

increases when the layers are thinner. This interpretation would also be consistent with the shape

of the Si-H stretching band of samples A1, A2 (figure 2) and B4 (figure 3). Whereas the three

avgsamples presented very similar average carbon contents, the sample B4 (x = 0.05), which

consists of 80 elementary pairs of layers, shows a more pronounced 2090 cm contribution than-1

avg avgthe samples A1 (x = 0.04) and A2 (x = 0.07), which respectively consist of 15 and 25

10

elementary pairs of layers.

4. Summary and conclusions

1-x xFrom the IRS analysis of three series of a-Si:H/a-Si C :H multilayers and the base

materials used in them, we can conclude that H is bonded to Si atoms only in monohydride

1-x xgroups. Although the H content is higher in a-Si C :H layers due to their porous structure, no

preference for Si or C atoms was detected. The H content in the multilayers correlates well with

their mean composition, showing an intermediate behavior between that of the two base

materials, and no additional H was detected at the interfaces.

An increase in the 2090 cm contribution of the Si-H stretching band was observed when-1

the average carbon content is increased, and also in samples with similar average carbon contents

when the number of interfaces is increased (i.e. the elementary layers are thinner). The analysis

of the absorption-free zone of the spectra of the multilayers with variable mean composition

allowed us to obtain their effective refraction index and to deduce the index of the base materials.

1-x xThe refraction index of the a-Si C :H layers with low thicknesses deduced in this way is higher

1-x xthan that deduced from homogeneous films. This suggests that the a-Si C :H layers in

multilayers present a denser structure. Thus, the rise in the 2090 cm contribution when the-1

number of interfaces is increased should be attributed to the Si-H bonds in the vicinity of C

atoms.

Acknowledgements

We acknowledge the collaboration of the Scientific-Technical Services of the Universitat

11

de Barcelona where IR measurements were performed. This work was supported by the DGICYT

of the Spanish Government under program PB89-0236.

12

References

[1] Y. Tawada, K. Tsuge, M. Kondo, H. Okamoto and Y. Hamakawa, J. Appl. Phys. 53

(1982) 5273.

[2] M. Yoshimoto and H. Matsunami, J. Appl. Phys. 67 (1990) 6301.

[3] R. Schwartz, T. Fischer, P. Hanesch, T. Muschik, J. Kolodzey, H. Cerva, H.L.

Meyerheim and B.M.U. Scherzer, Appl. Surf. Sci. 50 (1991) 456.

[4] N. Bernhard, M. Kirsch, R. Eigenschenk, M. Bollu, C. Wetzel, F. Müller, and R.

Schwarz, Mat. Res. Soc. Symp. Proc. 192 (1990) 237.

[5] M.H. Brodsky, M. Cardona and J.J. Cuomo, Phys. Rev. B 16 (1977) 3556.

[6] M. Cardona, Phys. Status Solidi (b) 118 (1983) 463.

[7] H. Wieder, M. Cardona and C.R. Guarneri, Phys. Status Solidi (b) 92 (1979) 99.

[8] F. Maass, J. Bertomeu, J.M. Asensi, J. Puigdollers, J. Andreu, J.C. Delgado and

J. Esteve, Appl. Surf. Sci. 70/71 (1993) 768.

[9] R.R. Koropecki, F. Alvarez and R. Arce, J. Appl. Phys. 69 (1991) 7805.

[10] J. Bertomeu, J.M. Asensi, J. Puigdollers, J. Andreu and J.L. Morenza, Vacuum 44 (1993)

129.

[11] H. Ugur, R. Johanson and H. Fritzsche. In: "Tetrahedrally-Bonded Amorphous Silicon

Semiconductors", Ed. D. Adler and H. Fritzsche (Plenum, New York, 1985), p. 425.

[12] J. Bertomeu, J. Puigdollers, J.M. Asensi and J. Andreu, J. Non-Cryst. Solids 164-166

(1993) 861.

13

Figure captions

Figure 1. Infrared transmittance spectra corresponding to the two base materials used in the

multilayers.

Figure 2. Infrared transmittance spectra corresponding to the multilayers of the series A,

which presented a variable mean composition and a constant thickness of the

1-x xelementary a-Si C :H layers. The first number indicates the number of elementary

pairs of layers, and the numbers parenthesized are respectivelly the thickness of

1-x xa-Si:H and a-Si C :H in Å.

Figure 3. Infrared transmittance spectra corresponding to the multilayers of the series B,

which presented a variable mean composition and a constant thickness of the

elementary a-Si:H layers. The convention in the description of the multilayers is the

same that in Fig. 2.

Figure 4. Hydrogen bonded to silicon density as a function of the average carbon content of

the multilayers with variable mean composition (series A and B). The same values

for the two base materials are plotted. The dashed line is just for connecting these

two points.

Figure 5. Effective index of refraction of the multilayers of series A as a function of the

Si SiCthickness of the a-Si:H layer. From the fit, the values of n and n are obtained.

Figure 6. Infrared transmittance spectra corresponding to the multilayers of the series C,

which presented a constant mean compostion. The convention in the description of

the multilayers is the same that in Fig. 2.

Figure 7. Absorption spectra of the Si-H stretching band corresponding to the multilayers of

the series C.

14

Si SiCTable I. Multilayers used in this study. d and d are respectively the thicknesses of a-Si:H and

1-x xa-Si C :H layers and N is the number of elementary pairs of layers.

Si SiCSample d (Å) d (Å) N

A1 263 37 15

A2 147 37 25

A3 73 37 40

A4 30 37 60

A5 18 37 80

A6 7 37 100

B1 55 205 20

B2 55 75 40

B3 55 32 60

B4 55 9 80

C1 156 195 15

C2 60 67 40

C3 30 35 80

15

16

17

18

19

20

21